MOSFET switch for VCC and GND - no additional driver power supply ( is it possible ? )

[andre_luis]

Hey guys,

I was tasked to develop a project where one of the customer's requirements is that not only the 24V have to be switched, but also the GND; this must be done with semiconductor components, no relay allowed due to vibration issues. I am not willing to question such a specification regarding disconnection from GND, I will implement it as stated. The general diagram is as follows:

It would be nice if it didn't need an extra power supply, however, the only topology that came to mind was the use of 2 N-channel MOSFETs driven by opto-coupler, as shown below:

Regarding VDD1, this could be obtained from the 24V itself, but the VDD2, I'm not sure if I'm missing any obvious detail such as the need to extract it via some bootstrap circuit, but would there be any impediment to obtain it from 24V as well? I mean, VDD1 and VDD2 connected to the 24V power supply ?

By the way, eventually the same circuit could be used with 48V as well, but that's an offtopic detail on this question for now.

 

[FvM]

MOSFET driven by photovoltaic opto coupler, e.g. Vishay VOM1271 can be on option. Need to specify switching speed.

 

[KlausST]

Hi,

If you disconnect GND as well as 24V, then this node is floating. So the voltages with respect to the remaining nodes is not defined.
You can't measure nor compare (comparator) "not defined" voltages.

Ignoring this..
You could combine all GNDs with diodes the negative pin of a capacitor (anode to capacitor).
and all 24V nodes with diodes to the positive pin of the capacitor (cathode to capacitor).
Then use the capacitor as your logic supply.

Klaus

 

[dick_freebird]

Are you against solid state charge pumps as well?
There are now many isolating MOSFET drivers and
some include their own gate drive voltage pump,
most would work with an external one.

The control scheme seems not to be shown. If it is
self contained (just working on "who's up, who's not?")
then presumably each 24V could power its own pump,
wire-or those for main power and do logiv on each to
see if it's up or not.

Or something like that.

For charge pump you might also consider the tinest
"aircraft 28V" input isolated output DC-DC module you
can find, times two or four, same idea. Like PICO has itty
bitty ones. Output voltage being whatever your MOSFET
drivers like for gate supply. Isolated outputs can give
you rails referred and spaced how you like. Crude diode
/ NPN logic can give you the enables.

 

[andre_luis]

Thanks for the answers:

@KlausST , I'm with you, I don't like disconnecting the GND, but this requirement has already come to me like that. Anyway, in the firmware I will switch both VCC and GND slightly out of sync, disconnecting GND last, and reconnecting it first - just enough to not violate the no-power dead time limit and to ensure no reference bus net fluctuation.

@dick_freebird , the idea of the charge pump is what I wanted before being aware of the solid state relay tip. I would never imagine that it would be possible to harvest the energy for the isolated part of the opto from the 'primary' stage; not that surprisingly anyway, it is required an insignificant energy to charge the Gate. The option made with discrete components I don't know how to implement it, more exactly, the circuit itself.

@FvM , this option is interesting, I had not previously worked with this type of optocoupler.
I made some estimates below to the switching time.
Am I doing it right?

Data:​

• MOSFET Gate charge = 160 nC
• OPTO Short circuit current = 25 μA

Calculation:​

• t = Q/I = 160nC / 25uA = ~6ms

MOSFET specs

OPTO specs

BTW: I chose another opto model, quite similar to the one you suggested, but with stock at Digikey and Mouser.

 

[crutschow]

Would this work for you (if you don't need high speed switching):
Note that the high side needs to have a P-MOSFET to turn it fully on unless you have a gate voltage that exceeds the supply voltage by 10V.
My MOSFET selections are arbitrary from the models I have.

 

[KlausST]

Hi,

To extend FvM's idea:

There are optoMOS relays, combining the mentioned opto driver with a MOSFET.
Do a search for "optoMOS" or "Mosfet relay".
They are available for DC load as well as AC load.

Klaus

 

[mtwieg]

You don't mention a spec on how much voltage the switch needs to withstand while open. I assume this might swing both positive and negative, in which case you should have two anti-series FETs in each switch. Depending on the magnitude of the voltage, charge pump circuits may or may not be feasible. If not, then your only options are a galvanically isolated power supply, or use photovoltaic isolators (PVIs) like FvM states above.

 

[andre_luis]

Hi,
@crutschow, many thanks for the simulation; however, I would prefer to use both MOSFETs from the same kind, prefereably the N-type - although not having strict switching speed requirements ( even a dozen miliseconds would meet the need ). At the purchasing stage it will be easier to deal with just one model, whose range of options is somewhat restricted for the available budget, the above 'switch' will work with ~30A, continous.

@all, the arrangement purposed on post #5, based on everyone's inputs, seems to solve the issue; I'm just curious to know if I took the correct approach to assess the switching time with the selected MOSFET+OPTO set. I would appreciate if someone could give some insight about the Calculation topic on that post.


Side note:

The topology on the first picure on post #1 stems from the customer's original SPDT relay based solution where only one pair of contacts on the same coil met the need, but due to extreme vibration requirements on the target application this had to be redesigned with a 'silicon-based' solution.​

 

[crutschow]

I would prefer to use both MOSFETs from the same kind, prefereably the N-type

Then you will need ether the charge-pump driver approach, or the photo-voltaic opto-coupler, to generate a gate voltage greater than the supply voltage to fully turn on the top MOSFET, since it is operating as a source-follower, with the gate voltage needing to be greater then the source volage by more than the MOSFET Vgs(th) value.

 

[andre_luis]

Then you will need ether the charge-pump driver approach, or the photo-voltaic opto-coupler, to generate a gate voltage greater than the supply voltage to fully turn on the top MOSFET, since it is operating as a source-follower, with the gate voltage needing to be greater then the source volage by more than the MOSFET Vgs(th) value.

Hi,
Regarding this requirement, since the opto-driver has no external power supply, I assume that the V_GS voltage is generated internally by some kind of 'photoelectric' charge pump mechanism; note that from the application note it is not required any extra power, nor does it suggest using a pullup to the V_D. Another detail is that in the absence of more details in the datasheet, as only the short-circuit current is mentioned for the isolated part of the opto, this must act more as a current source, I guess.

 

[KlausST]

Hi,

as the symbol shows: there are a bunch of photdiodes in series to generate the high (V_GS) voltage.
(Plus some circuitry for proper operation, like fast turn OFF.)

Again: there are ready to buy solutions.

Klaus

 

[crutschow]

since the opto-driver has no external power supply

What opto-driver?

 

[mtwieg]

@all, the arrangement purposed on post #5, based on everyone's inputs, seems to solve the issue; I'm just curious to know if I took the correct approach to assess the switching time with the selected MOSFET+OPTO set. I would appreciate if someone could give some insight about the Calculation topic on that post.

Basically when on, the PVI behaves as a current source (Isc) so long as it's below its open circuit voltage (Voc). So basically dt=dQ/Isc.
The turn-on delay will be Qg(th)/Isc
The switching time will be Qgd/Isc
After that, Vgs will continue to rise until it hits Voc, where it will stay. Can use the Q vs Vgs curve to find how long that is.

Turn off behavior depends greatly on the PVI you use. Many have fast turn-off circuitry, but the actual behavior of this is usually not documented well. But it's always faster than turn-on time. Best to investigate this with actual testing.

 

[cupoftea]

Connect both GNDs together with 1MEG, so they are kind of referenced the same. (or lower than 1MEG if you can get away with it.)

Then you need 4 back-to-back NFET switches......and get the drive voltage for these from the 4 isolated outputs of a 4_output flyback SMPS. Send the signal to switch them with 4 digitial isolators.

Beware inrush when you switch eg 24V to a non-ascerted "24v".

Put an RC snubber across the back2back fets.

--- Updated Mar 4, 2023 ---

Whatever you do, dont do "kissed_ON" FETs........they are not safe.
Ill send you a doc of what that is if you want...but i am sure you know.

--- Updated Mar 4, 2023 ---

SN6507 can also give you a cheap hi side drive supply if you want.

https://www.ti.com/lit/ds/symlink/sn6507.pdf?ts=1677936861439&ref_url=https%253A%252F%252Fwww.google.com%252F

 

[andre_luis]

What opto-driver?

Hi; This one: APV1111GV
Indeed, it was somewhat hidden on the quoted datasheet fragment at the end of the post #5.

The turn-on delay will be Qg(th)/Isc

Thank you, this matches with the approach posed on post #5; the central point then was to check if it is correct and if there were any other boundary condition to be considered.

Turn off behavior depends greatly on the PVI you use. Many have fast turn-off circuitry, but the actual behavior of this is usually not documented well. But it's always faster than turn-on time. Best to investigate this with actual testing.

The one selected do have this information. it lies within an acceptable range, 0,1ms.

 

[crutschow]

You do know that those cost over US$5 in unit quantities?

 

[cupoftea]

presumably these "Mosfet relays" that are popping up all over digikey are no good for you?